SaFi Bank Space : transaction_snapshot_history_features

[ Feature Explanation ] [ Snapshot Transaction Features ]

Last Updated: caa. 28 Dec 22

Following JSON details the fields that will be sent to both Slacker Manager. The source for the intrabank transactions historical features is retrieved from the following source.

Kafka Topic: transactions.transaction-snapshot.snapshot.v1
BigQuery Table: PROJECT_ENV.transactions.transactions_transaction_snapshot_snapshot_v1

Current method of retrieving the historical features, is to run an SQL query against the specified BigQuery table and aggregate the historical features as part of this query. A sample of the query execution will return the following table and values.

The keys of the table are:

cust_id - senderInfo.customerId
beneficiary_acc_id - beneficiaryInfo.customerId

Feature Explanation

1. Time Calculation and Definitions

Historical features are calculated as the difference in the number of hours between now and the expectedDate of transaction. Here one day is defined as 24 hours in the hour_diff variable.

Time is also calculated as a floating window for real-time queries. For instance, last 24 hours will be calculated from the current DATETIME.

DATETIME_DIFF(CURRENT_DATE(), SAFE_CAST(expectedDate as date), HOUR) as hour_diff

360 days: hour_diff <= 360*24
30 days: hour_diff <= 30*24
7 days: hour_diff <= 7*24
1 day: hour_diff <= 1*24

2. Transaction Status

All transaction status calculated for this featurestore has status = "SUCCESS" . All other transaction statuses are filtered out during the SQL query.

Snapshot Transaction Features

transaction_snapshot_history_features

Response Result	Description	feature_id	Type
cust_id	`entity_id`	`customerid`	UUID
beneficiary_acc_id	The recipient account id	`beneficiary_acc_id`	STRING
category	The category in which the given transaction belongs to	`category`	STRING
transactionType	The transaction type of the given transaction ID	`transactiontype`	STRING
direction	Indicates the direction of the transaction `DEBIT` - Outgoing `CREDIT` - Incoming	`direction`	STRING
status	The status of the transaction - E.g. PENDING, SUCCESS, etc.	`status`	STRING
tx_total_count	Total count of transactions made till date	`tx_total_count`	INT
unique_tx_days	Total unique number of days, in which a SUCCESSFUL transaction occured	`unique_tx_days`	INT
tx_count_1d	Total count of transactions made in the last 24 hours	`tx_count_1d`	INT
tx_sum_1d	Total amount of transactions made in the last 24 hours	`tx_sum_1d`	FLOAT
tx_avg_1d	Average amount of transactions made in the last 24 hours	`tx_avg_1d`	FLOAT
tx_count_7d	Total count of transactions made in the last 7 days	`tx_count_7d`	INT
tx_sum_7d	Total amount of transactions made in the last 7 days	`tx_sum_7d`	FLOAT
tx_avg_7d	Average amount of transactions made in the last 7 days	`tx_avg_7d`	FLOAT
tx_count_30d	Total count of transactions made in the last 30 days	`tx_count_30d`	INT
tx_sum_30d	Total amount of transactions made in the last 30 days	`tx_sum_30d`	FLOAT
tx_avg_30d	Average amount of transactions made in the last 30 days	`tx_avg_30d`	FLOAT
tx_count_360d	Total count of transactions made in the last 360 days	`tx_count_360d`	INT
tx_sum_360d	Total amount of transactions made in the last 360 days	`tx_sum_360d`	FLOAT
tx_avg_360d	Total amount of transactions made in the last 360 days	`tx_avg_360d`	FLOAT
tx_sum_last5_out	Total amount of the last 5 outgoing transactions	`tx_sum_last5_out`	FLOAT
tx_avg_last5_out	Average amount of the last 5 outgoing transactions	`tx_avg_last5_out`	FLOAT
tx_days_since_last	Number of days since the last transaction was made	`tx_days_since_last`	INT

SaFi Bank Space : transaction_snapshot_history_features

Feature Explanation

Snapshot Transaction Features

Attachments: